Font Size: a A A

Sentiment Analysis Of Shanghai Inbound Tourist During The Epidemic With Image-text Fusion Machine Learning ——Taking Instagram As An Example

Posted on:2022-11-10Degree:MasterType:Thesis
Country:ChinaCandidate:Y Y HuaFull Text:PDF
GTID:2518306773994399Subject:Tourism
Abstract/Summary:PDF Full Text Request
As social media networks' increasing growth,people spend more of their lives on social media,such as expressing their emotions and opinions through pictures,texts,videos and other forms.And software such as Instagram uses pictures as the main content of blog posts,while text is just an additional supplement.In order to make sentiment analysis more accurate,how to analyze sentiment in multimodal data is an opportunity and a challenge facing the current sentiment analysis field.In this article,the author fully introduces sentiment analysis technology and deep learning related technologies.After testing several model structures like VGG16,VGG19,Inception V3,Word2 Vec,Glo Ve,LSTM,and Bi-LSTM,the VGG16 and Bi-LSTM models were selected as the base models of the image-text sentiment analysis model.On this basis,a dropout layer was added,the settings of the output layer were modified,and different hyperparameter combinations were tried.Finally,the construction of the image-text fusion sentiment analysis model is completed by means of decisionlevel fusion,and have reached 75% accuracy on the test set.After proposing the sentiment analysis model of the image-text fusion,this paper takes this as the core,and designs a set of tourist emotion analysis framework from three perspectives: tourists' feature,tourist emotional attitudes,and tourist's interest points.Subsequently,based on Trip Advisor.com's ranking of Shanghai tourist attractions,this paper obtained corpus data about the attractions on Instagram.After a series of data cleaning and sorting operations,taking the data as an example,the analysis of tourist satisfaction was carried out,including the characteristics of tourists,the distribution of tourists' emotional attitudes between scenic spots and different time periods,the sentiment differences within different group of tourists such as influence or nationality,the distribution of tourist interest points in each scenic spot,and the Tourist satisfaction ranking of scenic spots.In the end,comprehensive suggestions such as "focusing on the tourism needs of emerging developing countries" and "adjusting development strategies according to the type of scenic spots" and specific suggestions for each scenic spot were given.
Keywords/Search Tags:Image-text fusion, Deep Learning, Sentiment Analysis, Inbound Tourism
PDF Full Text Request
Related items